HIT3002: Introduction to Artificial Intelligence

Transcription

1 HIT3002: Introduction to Artificial Intelligence Intelligent Agents Outline Agents and environments. The vacuum-cleaner world The concept of rational behavior. Environments. Agent structure. Swinburne University of Technology 1

2 Agents and environments Agents include human, robots, softbots, thermostats, etc. The agent function maps percept sequence to actions An agent can perceive its own actions, but not always it effects. f : P* A Agents and environments The agent function will internally be represented by the agent program. The agent program runs on the physical architecture to produce f. Swinburne University of Technology 2

3 The vacuum-cleaner world An example Environment: squares A and B Percepts: [location and content] e.g. [A, Dirty] Actions: left, right, suck, and no-op The vacuum-cleaner world Agent function Percept sequence [A,Clean] [A, Dirty] [B, Clean] [B, Dirty] [A, Clean],[A, Clean] [A, Clean],[A, Dirty] Action Right Suck Left Suck Right Suck Swinburne University of Technology 3

4 The vacuum-cleaner world An agent program function REFLEX-VACUUM-AGENT ([location, status]) return an action if status == Dirty then return Suck else if location == A then return Right else if location == B then return Left What is the right function? Can it be implemented in a small agent program? The concept of rationality A rational agent is one that does the right thing. Every entry in the table is filled out correctly. What is the right thing? Approximation: the most successful agent. Measure of success? Performance measure should be objective E.g. the amount of dirt cleaned within a certain time. E.g. how clean the floor is. Performance measure according to what is wanted in the environment instead of how the agents should behave. Swinburne University of Technology 4

5 Rationality What is rational at a given time depends on four things: Performance measure, Prior environment knowledge, Actions, Percept sequence to date (sensors). DEF: A rational agent chooses whichever action that maximizes the expected value of the performance measure given the percept sequence to date and prior environment knowledge. Rationality Rationality omniscience An omniscient agent knows the actual outcome of its actions. Rationality perfection Rationality maximizes expected performance, while perfection maximizes actual performance. Swinburne University of Technology 5

6 Rationality The proposed definition requires: Information gathering/exploration To maximize future rewards Learn from percepts Extending prior knowledge Agent autonomy Compensate for incorrect prior knowledge Is the vacuum cleaner agent rational? Depend! For example, it s rational under the following assumptions: Performance measure: 1 point for each clean square over lifetime of 1000 steps geography known but dirt distribution, initial position of agent not known Clean squares stay clean, sucking cleans squares Left and Right don t take agent outside environment Available actions: Left, Right, Suck, NoOp Agent knows where it is and whether that location contains dirt Swinburne University of Technology 6

7 Environments To design a rational agent we must specify its task environment. PEAS description of the environment: Performance Environment Actuators Sensors Environments E.g. Fully automated taxi: PEAS description of the environment: Performance Safety, destination, profits, legality, comfort Environment Streets/freeways, other traffic, pedestrians, weather, Actuators Steering, accelerating, brake, horn, speaker/display, Sensors Video, sonar, speedometer, engine sensors, keyboard, GPS, Swinburne University of Technology 7

8 Environment types Environment types Fully vs. partially observable: an environment is full observable when the sensors can detect all aspects that are relevant to the choice of action. Swinburne University of Technology 8

9 Environment types Fully vs. partially observable: an environment is full observable when the sensors can detect all aspects that are relevant to the choice of action. Environment types Deterministic vs. stochastic: if the next environment state is completely determined by the current state the executed action then the environment is deterministic. Swinburne University of Technology 9

10 Environment types Deterministic vs. stochastic: if the next environment state is completely determined by the current state the executed action then the environment is deterministic. Environment types Episodic vs. sequential: In an episodic environment the agent s experience can be divided into atomic steps where the agents perceives and then performs A single action. The choice of action depends only on the episode itself Swinburne University of Technology 10

11 Environment types Episodic vs. sequential: In an episodic environment the agent s experience can be divided into atomic steps where the agents perceives and then performs A single action. The choice of action depends only on the episode itself Environment types Static vs. dynamic: If the environment can change while the agent is choosing an action, the environment is dynamic. Semi-dynamic if the agent s performance changes even when the environment remains the same. Swinburne University of Technology 11

12 Environment types Static vs. dynamic: If the environment can change while the agent is choosing an action, the environment is dynamic. Semi-dynamic if the agent s performance changes even when the environment remains the same. SEMI Environment types Discrete vs. continuous: This distinction can be applied to the state of the environment, the way time is handled and to the percepts/actions of the agent. SEMI Swinburne University of Technology 12

13 Environment types Discrete vs. continuous: This distinction can be applied to the state of the environment, the way time is handled and to the percepts/actions of the agent. SEMI Environment types Single vs. multi-agent: Does the environment contain other agents who are also maximizing some performance measure that depends on the current agent s actions? SEMI Swinburne University of Technology 13

14 Environment types Single vs. multi-agent: Does the environment contain other agents who are also maximizing some performance measure that depends on the current agent s actions? SEMI Environment types The simplest environment is Fully observable, deterministic, episodic, static, discrete and single-agent. Most real situations are: Partially observable, stochastic, sequential, dynamic, continuous and multi-agent. Swinburne University of Technology 14

15 Agent types How does the inside of the agent work? Agent = architecture + program All agents have the same skeleton: Input = current percepts Output = action Program= manipulates input to produce output Note difference with agent function. Agent types Function TABLE-DRIVEN_AGENT(percept) returns an action static: percepts, a sequence initially empty table, a table of actions, indexed by percept sequence append percept to the end of percepts action LOOKUP(percepts, table) return action This approach is doomed to failure Swinburne University of Technology 15

16 Agent types Four basic kind of agent programs will be discussed: Simple reflex agents Model-based reflex agents Goal-based agents Utility-based agents All these can be turned into learning agents. Agent types; simple reflex Select action on the basis of only the current percept. E.g. the vacuum-agent Large reduction in possible percept/action situations(next page). Implemented through condition-action rules If dirty then suck Swinburne University of Technology 16

17 The vacuum-cleaner world function REFLEX-VACUUM-AGENT ([location, status]) return an action if status == Dirty then return Suck else if location == A then return Right else if location == B then return Left Reduction from 4 T to 4 entries Agent types; simple reflex function SIMPLE-REFLEX-AGENT(percept) returns an action static: rules, a set of condition-action rules state INTERPRET-INPUT(percept) rule RULE-MATCH(state, rule) action RULE-ACTION[rule] return action Will only work if the environment is fully observable otherwise infinite loops may occur. Swinburne University of Technology 17

18 Agent types; reflex and state To tackle partially observable environments. Maintain internal state Over time update state using world knowledge How does the world change. How do actions affect world. Model of World Agent types; reflex and state function REFLEX-AGENT-WITH-STATE(percept) returns an action static: rules, a set of condition-action rules state, a description of the current world state action, the most recent action. state UPDATE-STATE(state, action, percept) rule RULE-MATCH(state, rule) action RULE-ACTION[rule] return action Swinburne University of Technology 18

19 Agent types; goal-based The agent needs a goal to know which situations are desirable. Things become difficult when long sequences of actions are required to find the goal. Typically investigated in search and planning research. Major difference: future is taken into account Is more flexible since knowledge is represented explicitly and can be manipulated. Agent types; utility-based Certain goals can be reached in different ways. Some are better, have a higher utility. Utility function maps a (sequence of) state(s) onto a real number. Improves on goals: Selecting between conflicting goals Select appropriately between several goals based on likelihood of success. Swinburne University of Technology 19

20 Agent types; learning All previous agentprograms describe methods for selecting actions. Yet it does not explain the origin of these programs. Learning mechanisms can be used to perform this task. Teach them instead of instructing them. Advantage is the robustness of the program toward initially unknown environments. Agent types; learning Learning element: introduce improvements in performance element. Critic provides feedback on agents performance based on fixed performance standard. Performance element: selecting actions based on percepts. Corresponds to the previous agent programs Problem generator: suggests actions that will lead to new and informative experiences. Exploration vs. exploitation Swinburne University of Technology 20

21 Summary: Agents An agent perceives and acts in an environment, has an architecture, and is implemented by an agent program. Task environment PEAS (Performance, Environment, Actuators, Sensors) An ideal agent always chooses the action which maximizes its expected performance, given its percept sequence so far. An autonomous learning agent uses its own experience rather than built-in knowledge of the environment by the designer. An agent program maps from percept to action and updates internal state. Reflex agents respond immediately to percepts. Goal-based agents act in order to achieve their goal(s). Utility-based agents maximize their own utility function. Representing knowledge is important for successful agent design. The most challenging environments are not fully observable, nondeterministic, dynamic, and continuous Swinburne University of Technology 21